BRPI0707135A2 - apparatus and method for signal coding and decoding - Google Patents

Abstract

APPARATUS AND METHOD FOR SIGNAL CODING AND DECODING. Encoding and decoding devices and methods for encoding and decoding are provided. The decoding method includes extracting a plurality of encoded signals from a continuous stream of input bits, determining which of a plurality of decoding methods should be used to decode each of the encoded signals, decoding the encoded signals using the determined decoding methods and synthesize the decoded signals. In this way, it is possible to encode signals with different characteristics at an ideal bit rate by classifying the signals into one or more classes according to the characteristics of the signals and it is possible to encode each of the signals using a coding unit that can best serve to the class to which a corresponding signal belongs. Furthermore, it is possible to efficiently encode various signals including audio and speech signals.

Description

"APPARATUS AND METHOD FOR SIGNAL coding and coding"

TECHNICAL FIELD

The present invention relates to decoding and decoding apparatus and methods for encoding and decoding, and more particularly to coding and decoding apparatus and methods for encoding and decoding which can encode or decode signals at an optimal bit rate according to the characteristics of the signs.

BACKGROUND OF THE INVENTION

Conventional audio encoders can provide high quality audio signals at a high rate of 48 kbps or higher, but are inefficient for speech processing. Conversely, conventional speech encoders can effectively encode speech signals at a low bit rate of 12 kbps or less, but are insufficient to encode various audio signals.

DISCLOSURE OF INVENTION

TECHNICAL PROBLEM

The present invention provides encoding and decoding apparatus and methods for encoding and decoding which can encode or decode signals (e.g., speech and audio signals) with different characteristics at an optimal bit rate. .

TECHNICAL SOLUTION

According to one aspect of the present invention, there is provided a method for decoding which includes extracting a plurality of encoded signals from a continuous stream of input bits, determining which of a plurality of decoding methods should be used to decode each of the encoded signals, to decode the signals. encoded using the decoding methods determined to synthesize the decoded signals.

According to another aspect of the present invention, a decoding apparatus is provided which includes a bit unwrapping module that extracts a plurality of encoded signals from a continuous stream of input bits, a module for determining decoder which determines which of a plurality of decoding units should be used to decode each of the encoded signals, a decoding module that includes the decoding units and decodes the encoded signals using the determined decoding units, and a sync module. -testing that synthesizes the decoded signals.

In accordance with another aspect of the present invention, there is provided a method for encoding which includes splitting an input signal into a plurality of divided signals, determining which of a plurality of co-diffusion methods should be used. To encode each of the split signals based on the characteristics of each of the split signals, encode the split signals using the determined encoding methods, and generate a continuous stream debits based on the encoded split signals.

According to another aspect of the present invention, there is provided a coding apparatus including a signal splitting module that splits an input signal into a plurality of divided signals, an encoder determining module which determines which of a A plurality of coding units shall be used to encode each of the split signals based on the characteristics of each of the split signals, a coding module which includes coding units and coding the divided signals using the given coding units, and a bit wrapping module which generates a continuous stream of bits based on the coded split signals.

ADVANTABLE EFFECTS

In this way, it is possible to encode signals with different characteristics at an optimal bit rate by classifying the signals into one or more classes according to the characteristics of the signals, and encode each signal using a coding unit that can best suit the class at the same time. which a corresponding signal belongs. Moreover, it is possible to efficiently encode various signals, including audio and speech signals.

BRIEF DESCRIPTION OF DRAWINGS

Figure 1 is a block diagram of a coding apparatus in accordance with one embodiment of the present invention;

Fig. 2 is a block diagram of one embodiment of a classification module illustrated in Fig. 1;

Figure 3 is a block diagram of a embodiment of a preprocessing unit illustrated in Figure 2. Figure 4 is a block diagram of an apparatus for calculating the perceptual entropy of an input signal according to one embodiment. of the present invention;

Fig. 5 is a block diagram of another embodiment of the classification module shown in Fig. 1;

Fig. 6 is a block diagram of one embodiment of a signal division unit shown in Fig. 5;

Figures 7 and 8 are diagrams for explaining methods for merging a plurality of split signals in accordance with embodiments of the present invention;

Figure 9 is a block diagram of another mode of the signal splitting unit shown in Figure 5;

Figure 10 is a diagram for explaining a method for dividing an input signal into a plurality of divided signals according to one embodiment of the present invention;

Fig. 11 is a block diagram of one embodiment of a unit of determination illustrated in Fig. 5;

Fig. 12 is a block diagram of one embodiment of a coding unit illustrated in Fig. 1;

Figure 13 is a block diagram of another embodiment of the coding unit shown in Figure 1;

Fig. 14 is a block diagram of a coding apparatus according to another embodiment of the present invention;

Figure 15 is a block diagram of a decoding apparatus according to one embodiment of the present invention; and

Figure 16 is a block diagram of one embodiment of a synthesizer unit shown in Figure 15.

BEST MODE FOR CARRYING OUT THE INVENTION

In the following, the present invention will be described more fully with reference to the accompanying drawings, in which exemplary embodiments of the invention are shown.

Figure 1 is a block diagram of a coding apparatus in accordance with one embodiment of the present invention. Referring to Figure 1, the coding apparatus includes a rating module 100, a coding module 200 and a bit wrapping module 300.

The coding module 200 includes a plurality of the first through m-th coding units 210 and 220 which performs different coding methods.

The classification module 100 splits an input signal into a plurality of split signals and each of the split signals has one of the first to the most coding units 210 and 220. Some of the first to the most coding units 210 and 220 may be married to two or more divided signals or undivided signals.

The rating module 100 may allocate a number of bits to encode each of the divided signals or may determine the order in which the divided signals shall be encoded.

The coding module 200 encodes each split signal using any of the first up to m-coding units 210 and 220 which is matched to a corresponding split signal. The rating module 100 analyzes the characteristics of each of the divided signals and chooses one of the first to mth coding units 210 and 220 which can more efficiently encode each divided signal according to the results of the analysis.

A coding unit that can more efficiently encode a divided signal can be considered capable of achieving higher compression efficiency.

For example, a divided signal that can easily be modeled as a coefficient and a residue can be effectively coded by a speech coder, and a divided signal that cannot be easily modeled as a coefficient and a residue can be efficiently coded by a audio encoder.

If the ratio of the energy of a residue obtained by modeling a divided signal to the divided signal energy is less than a predefined limit, the divided signal can be considered a signal that can be easily modeled.

Since a divided signal that exhibits a high redundancy on a time axis can be well modeled using a linear prediction method in which an actual signal is predicted based on an earlier signal, it can be more efficiently encoded by a do it using a linear prognostic coding method.

Bit wrapping module 300 generates a continuous stream of bits to be transmitted based on coded split signals provided by coding module 200 and additional coding information which considers coded split signals. Bit packaging module 300 can generate a continuous stream of bits with a variable bit rate using a flat bit method or a sliced arithmetic coding method.

Divided signals or bandwidths that are not coded due to bit rate constraints may be restored from decoded signals or bandwidths provided by a decoder using an interpolation, extrapolation or replication method. Also, compensation information that considers split signals that are not encoded can be included in a continuous bit stream to be transmitted.

Referring to Figure 1, the classification module 110 may include a plurality of the first to nth grading units 110 and 120. Each of the first to the nth grading units 110 and 120 may divide the input signals into a plurality. split signal, convert a domain of the input signal, extract the characteristics of the input signal, classify the input signal according to the characteristics of the input signal, or match the input signal with one of the first to m-es coding units 210 and 220.

One of the first through nth grading units 110 and 120 may be a preprocessing unit which performs a preprocessing operation on the input signal so that the input signal can be converted into an efficiently encoding signal. The preprocessing unit may divide the input signal into a plurality of components, for example, an efficient component and a signal component, and may perform a preprocessing operation on the input signal before other rating units perform their operations.

The input signal may be preprocessed selectively according to the characteristics of the input signal, according to external environmental factors and according to the target bit rate, and only part of a plurality of split signals obtained from the input signal. input signal can be selectively preprocessed.

The rating module 100 may classify the input signals according to the perceptual characteristic information of the input signal provided by a psychoacoustic modeling module 400. Examples of the perceptive characteristic information include a masking limit, a ratio Signal Mask (SMR) and Perceptual Entropy.

In other words, the classification module 100 may divide the input signal into a plurality of split signals or may match each of the split signals with one or more of the first to mth coding units 210 to 220 according to the perceptual characteristic information. -activating the input signal, for example a masking limit and an SNR of the input signal.

In addition, the rating module 100 can receive information such as tone, zero crossing rate (ZCR) and input signal linear prognostic coefficient, and previous frame rating information, and can classify the input signal. according to the information received.

Referring to Figure 1, the coded result information transmitted by the coding module 200 may be fed back to the rating module 100.

Since the input signal is divided into a plurality of signals divided by the classification module 100e, it is determined by which of the first to mth decoding units 210 and 220, by what amount of bits, and in what order the divided signals. must be encoded, the divided signals are encoded according to the results of the determination. An amount of bits actually used to encode each of the divided signals may not necessarily be the same amount of bits allocated by the rating module 100.

Information that specifies the difference between the actual amount of bits used and the amount of bits allocated can be fed back to the rating module 100 so that the rating module 100 can increase the amount of bits allocated to other split signals. If the amount of bits actually used is larger than the allocated bit size, the rating module 100 can reduce the amount of bits allocated to other split signals.

A coding unit that actually encodes a split signal may not necessarily be the same coding unit that is matched to the split signal by rating module 100. In this case, information may be fed back into rating module 100, indicating that a unit coding that actually encodes a divided signal is different from a coding unit matched by the signal divided by the rating module 100. Then, the rating module 100 can match the divided signal with a different coding unit of the coding unit previously married to the divided signal.

The rating module 100 may further divide the input signal into a plurality of divided signals according to the newly encoded result information and feedback. In this case, the classification module 100 can obtain a plurality of split signals with a different structure than the previously obtained split signals.

If a coding operation chosen by the rating module 100 differs from a coding operation that is actually performed, information that considers the differences between them may be fed back to the rating module 100 so that the rating module 100 can redetermine the information. related to the coding operation.

Fig. 2 is a block diagram of one embodiment of a rating module 100 illustrated in Fig. 1. Referring to Figure 2, the first sorting unit may be a preprocessing unit that performs a preprocessing operation on an input signal so that the input signal can be effectively encoded.

Referring to Figure 2, the first classification unit 110 may include a plurality of the first to the 11th preprocessors 111 and 112 which perform different preprocessing methods. The first rating unit 110 may use one of the first through nth preprocessors 111 and 112 to preprocess an input signal according to input signal characteristics according to environmental factors. and according to a target bitrate. Also, the downgrading unit 110 may perform two or more preprocessing operations on the input signal using the first through 11th preprocessors 111 and 112.

Figure 3 is a block diagram of a modality of the first through nth preprocessors 111 and 112 shown in Figure 2. Referring to Figure 3, a preprocessor includes a coefficient puller 113 and a waste extractor 114.

The coefficient puller 113 analyzes an input signal and extracts from the input signal a coefficient that represents the characteristics of the input signal. Residue extractor 114 extracts from the input signal a residue with redundant components removed using the extracted coefficient.

The preprocessor can perform a linear prognostic decoding operation on the input signal. In this case, the coefficient extractor 113 extracts a linear prognostic coefficient from the input signal by performing linear prognostic analysis on the input signal, and the residual extractor 114 extracts a residue from the input signal using the linear prognostic coefficient provided by the input extractor. coefficient 113. The residual life-span redundancy may have the same shape as white noise.

In the following, a li-near prognostic analysis method according to one embodiment of the present invention will be described in detail.

A predicted signal obtained by linear deprognostic analysis can be composed of a li-near combination of the previous input signals, as indicated by Equation (1):

Mathematical value 1

<formula> formula see original document page 13 </formula>

where ρ indicates a linear prognostic order, ιaté ρ indicates linear prognostic coefficients that are obtained by minimizing an average square error (MSE) between an input signal and an estimated signal.

A transfer function P (z) for linear deprognostic analysis can be represented by Equation (2):

Mathematical value 2In relation to Figure 3, the preprocessor can extract a linear prognostic coefficient and a residual of an input signal using a distorted linear prognostic (WLPC) coding method, which is another type of a -5 linear prognostic analysis. The WLPC method can be performed by replacing a Z-1 unit delay with an all-pass filter with a transfer function A (z). Transfer function A (z) can be represented by E-quation (3):

Mathematical value 3

<formula> formula see original document page 14 </formula>

where it indicates the coefficient of a filter passed all. By varying the filter coefficient passes everything, it is possible to vary the resolution of a signal to be analyzed. For example, if a signal to be analyzed is highly concentrated in a certain frequency band, for example, if the signal to be analyzed is an audio signal that is highly concentrated in a low frequency band, the signal to be analyzed can be encoded. efficiently by setting the efficient pass everything goes so that the resolution of low frequency band signals can increase.

In the WLPC method, low frequency signals are analyzed with higher resolution than high frequency signals. Thus, the WLPC method can achieve high prognostic performance for low frequency signals and can better model low frequency signals.

The pass-all coefficient may vary along a geometric axis of time according to the characteristics of an input signal, according to external environmental factors and according to a target bit rate. If the coefficient passes everything varies over time, an audio signal obtained by decoding can be considerably distorted. Thus, when the pass-all coefficient varies, a smoothing method can be applied to the pass-all coefficient so that the pass-all coefficient can vary gradually, and that signal distortion can be minimized. The devalue range that can be determined as a current pass-all coefficient value can be determined by previous all-pass coefficient values.

A masking threshold, instead of an original signal, can be used as an input for estimating a linear prognostic coefficient. More specifically, a masking threshold can be converted to a time domain signal, and WLPC can be performed using the time domain signal as an input. The prognosis of a linear prognostic coefficient can be further accomplished using a residue as an input. In other words, linear prognostic analysis may be performed more than once, thereby obtaining an additionally whitened residue.

Referring to Figure 2, the first classification unit 110 may include a first preprocessor 111 which performs the above-described linear prognostic analysis with respect to Equations (1) and (2) and a second preprocessor (not shown) which performs the WLPC. The first declassification unit 100 may choose one from the first processor11 and the second preprocessor, or may decide not to perform linear prognostic analysis on an input signal according to the characteristics of the input signal according to the characteristics of the input. external environmental factors and according to a target bit rate.

If the pass-all coefficient has a value of 0, the second preprocessor may be the same as the first preprocessor 111. In this case, the first rating unit 110 may include only the second preprocessor and choose one of the analysis method. prediction method and WLPC method, according to the value of the passed coefficient. Also, the first classification unit 110 may perform linear prognostic analysis or any of the linear prognostic analysis method and the WLPC method is chosen in frame units.

Information indicating whether linear prognostic analysis should be performed and information indicating which of the linear prognostic analysis method and WLPC methods is chosen can be included in a continuous stream of bits to be transmitted.

Bit packaging module 300 receives from the first classification unit 110 a linear prediction coefficient, information indicating whether linear prediction coding should be performed and information indicating a linear prediction coder that is actually used. Then the bit marshaling module 300 inserts all information received into a stream of bits to be transmitted.

The amount of bits required to encode an input signal into a signal with a sound quality almost indistinguishable from that of the original input signal can be determined by calculating the input dosing perceptual entropy.

Figure 4 is a block diagram of an apparatus for calculating perceptual entropy according to one embodiment of the present invention. Referring to Figure 4, the apparatus includes a filter bank 115, a linear prognostic unit 116, a psychoacoustic modeling unit 117, a first bit calculation unit 118 and a second bit calculation unit 119.

PE perceptual entropy of an input signal can be calculated using Equation (4):

Mathematical value 4:

<formula> formula see original document page 17 </formula>

where X (ejw) affects the energy level of the original input signal, and T (ejw) indicates a masking limit. .

In a WLPC method that involves the use of an all-pass filter, the perceptual entropy of an input signal can be calculated using the proportion of the energy of an input signal residue by a residue masking threshold. More specifically, a coding apparatus using the WLPC method can calculate the perceptual entropy PE of an input signal using Equation (5):

Mathematical value 5

<formula> formula see original document page 18 </formula>

where R (Qjw) indicates the energy of a input signal residue and T (Qjw) indicates a masking limit of a residue.

The masking limit T (Qjw) can be represented by Equation (6):

Mathematical value 6

<formula> formula see original document page 18 </formula>

where T (Qjw) indicates a masking limit of an original signal and H (Qjw) indicates a transfer function for WLPC. The psychoacoustic modeling unit 320 can calculate the masking limit T (Qjw) using the T oversize limit (Qjw) in a scaling factor band domain and using the H (Qjw) transfer function.

Referring to Fig. 4, the first bit calculation unit 118 receives a residue obtained by the WLPC realized by the linear predictor unit 116 and an over-limit transmitted by the psycho-acoustic modeling unit 117. The filter bank 116 It can perform frequency conversion on an original signal, and the result of frequency conversion can be entered into the acoustic-psycho-modeling unit 117 and the second bit calculation unit119. Filter bank 115 can perform the deFourier transform on the original signal.

The first bit computation unit 118 can calculate perceptual entropy using the ratio of an original signal masking limit divided by a spectrum of a WLPC synthesis filter transfer function by the energy of the residue.

WPE distorted perceptual entropy of a signal that is divided into 60 or more non-uniform partition bands with different bandwidths can be calculated using WLPC, as indicated by Equation (7):

Mathematical value 7

<formula> formula see original document page 19 </formula>

where b indicates an index of a partition band obtained using a psychoacoustic model, eres (b) indicates the summed energies of the residuals in the partition band b, w_low (b) and w_high (b) respectively indicate lower and higher frequencies in the partition band b, Hbunear (w) indicates a limitede masking of a linearly mapped partition band, h (w) 2 indicates a linear deprognostic encoding energy spectrum (LPC) of a frame and nbres (w) indicates a limit of linear masking corresponding to the residue.

On the other hand, the distorted perceptual entropyWPEsub of a signal that is divided into 60 or more uniform department bands with the same bandwidth can be calculated using WLPC, as indicated by Equation (8): Mathematical value 8

<formula> formula see original document page 20 </formula>

where s indicates an index of a linearly partitioned subband, slow (w) and shigh (w) respectively indicate lower and higher frequencies in the closely partitioned subbands s, nbsub (s). ) indicates a masking limit of the linearly partitioned subband s, and Gsub (s) indicates the energy of the linearly partitioned subband s, that is, the sum of the frequencies in the linearly partitioned subband s. The nbsub masking threshold (s) is a minimum of a plurality of masking thresholds in the thinly partitioned subband.

Perceptual entropy may not be calculated for bands with the same bandwidth and higher limits than the sum of the input spectra. Thus, the WPEsab distorted perceptual entropy of Equation (8) may be smaller than the WPE distorted perceptual entropy of Equation (7), which provides high resolution for low frequency bands. bandwidth using WLPC, as indicated by Equation (9):

Mathematical value 9

<formula> formula see original document page 21 </formula>

where / indicates an index of a scaling factor band, nbSf (f) indicates a minimum masking limit of the scaling factor f, WPEsf indicates the proportion of an input of the scaling factor band / by a limitede masking of the scaling factor. scaling factor band /, and sp (s) - indicates the sum of all frequencies in the scaling factor band /, ie the energy of the scaling factor band /.

Fig. 5 is a block diagram of another embodiment of the rating module 100 shown in Fig. 1. Referring to Fig. 5, a rating module includes a signal division unit 121 and a determination unit 122.

More specifically, the signal division unit 121 divides an input signal into a plurality of divided signals. For example, signal division unit 121 may divide the input signal into a plurality of frequency bands using a subband filter. Frequency bands can have the same bandwidth or different bandwidths. As explained, a split signal may be encoded separately from the other divided signals by a coding unit which may best serve the characteristics of the split signal.

Signal division unit 121 may divide the input signals into a plurality of divided signals, for example a plurality of band signals, so that interference between the band signals can be minimized. Signal splitting unit 121 may have a dual filter bank structure. In this case, the signal division unit 121 may further divide each of the divided signals.

Division information considering the split signals obtained by signal division unit 121, for example, the total number of split signals and information from each of the split signals can be included in a continuous stream of bits to be transmitted. . A decoding apparatus can decode the split signal separately and synthesize the decoded signals relative to the split information, thereby restoring the original input signal.

Division information can be stored as a table. A stream of bits may include identifying information from a table used to split the original input signal.

The importance of each of the split signals (for example, a plurality of frequency band signals) for sound quality can be determined, and the rate of debits can be adjusted for each of the split signals according to the results of the determination. More specifically, the importance of a divided signal can be defined as a fixed value or as a non-fixed value that varies according to the characteristics of a frame input signal.

If the speech and audio signals are mixed into the input signal, the signal splitting unit 121 may divide the input signal into a speech signal and an audio signal according to the characteristics of the speak and audio signals. according to the characteristics of the audio signals.

The determining unit 122 may determine which of the first to mth coding units 210 and 220 coding code 200 may more efficiently encode one of the divided signals.

Determination unit 122 classifies the divided signals into numerous groups. For example, the terminating unit 122 may classify the signals into N classes and may determine which of the first through mth coding units 210 and 220 should be used to encode each signal divided by the matching of each one of the N classes with one of the first through m-th coding units 210 and 220.

More specifically, since the coding module 200 includes the first to mth coding units 210 and 220, the determining unit 122 can classify the signals divided into the first to mth classes, which can be encoded most efficiently by the first to the nth coding units 210 and 220, respectively.

For this, the characteristics of the signals that can be encoded most efficiently by each of the first to the mth coding units 210 and 220 may be determined in advance, and the characteristics of the first to the mth class may be defined according to the requirements of the present invention. -results of determination. Subsequently, the terminating unit 122 can extract the characteristics of each of the divided signals and classify each of the divided signals into one of the first to the mth classes that share the same characteristics as a corresponding divided signal. according to the extraction results.

Examples of the first through mth classes include a speech-to-speech class, a speechless class, a background noise class, a silence class, a tonal audio class, a non-tonal audio class, and a speech mix class. in voice / audio.

The determining unit 122 may determine which of the first to mth coding units 210 and 220 should be used to encode each of the divided signals by the perceptual characteristic information reference which considers the divided signals provided by the psychoacoustic modeling module 400, for example, the limits masking, SMRs or perceptual entropy levels of split signals.

The determining unit 122 may determine a number of bits to encode each of the divided signals or to determine the order in which the divided signals are to be encoded by reference to the perceptual characteristic information that considers the divided signals.

The information obtained by the determination performed by the determination unit 122, for example, indicates by which of the first to the mth coding units 210 and 220 and by what amount of bits each of the divided signals must be encoded and information indicates the order. in which the divided signals must be encoded, may be included in a continuous stream of bits to be transmitted.

Fig. 6 is a block diagram of one embodiment of signal division unit 122 illustrated in Fig. 5. Referring to Figure 6, a signal division unit includes a divider 123 and a merger 124.

Splitter 123 may divide an input signal into a plurality of divided signals. The merger 124 can merge split signals with similar characteristics into a single signal. For this, the blender 124 may include a synthesis filter bank.

For example, splitter 123 can split an input signal into 256 bands. Of the 256 bands, those with similar characteristics can be merged into a single merger bandap 124.

Referring to Fig. 7, the merger 124 may even have a plurality of split signals that are adjacent to each other in a single merged signal. In this case, merge 124 may merge a plurality of adjacent split signals into a single merged signal according to a predefined rule without regard to the characteristics of adjacent split signals.

Alternatively, with respect to Fig. 8, the merger 124 may merge a plurality of split signals with similar characteristics into a single merged signal, regardless of whether the split signals are adjacent to each other. In this case, the merger 124 may merge a plurality of split signals that can be efficiently encoded by the same coding unit into a single merged signal.

Fig. 9 is a block diagram of another mode of signal division unit 121 illustrated in Figure 5. Referring to Fig. 9, a desinal division unit includes a first divider 125, a second divider 126, and a third divider 127.

More specifically, the signal division unit 121 may hierarchically divide an input signal.

For example, the input signal may be divided into two signals divided by the first divider 125, one of the two divided signals may be divided into three signals divided by the second divider 126 and one of the three divided signals may be divided into three signals divided by the third divider127 . In this way, the input signal can be divided into a total of 6 divided signals. The signal division unit 121 may hierarchically divide the input signal into a plurality of bands with different bandwidths.

In the embodiment illustrated in Figure 9, an input signal is divided according to a 3-level hierarchy, but the present invention is not restricted to it. In other words, an input signal may be divided into a plurality of signals divided according to a hierarchy of 2 levels or 4 or more levels.

One of the first to third splitters 125 to 127 in signal division unit 121 may divide an input signal into a plurality of time domain signals.

Fig. 10 explains an embodiment of splitting an input signal into a plurality of signals divided by signal splitting unit 121.

In general, speech or audio signals are stationary for a short frame length period. However, sometimes speech or audio signals may have non-stationary characteristics, for example, during a transitional period.

In order to effectively analyze non-stationary signals and to improve the coding efficiency of such non-stationary signals, the coding apparatus according to the present embodiment may use an empirical mode wavelength or decomposition (EMD) method. In other words, the coding apparatus according to the present embodiment may analyze the characteristics of an input signal using an unfixed transform function. For example, signal division unit 121 may divide an input signal into a plurality of bands with varying bandwidths using a non-fixed frequency band subband filtering method. Here is a method for splitting an input signal into a plurality of signals divided by EMD will be described in detail.

In the EMD method, an input signal can be decomposed into one or more intrinsic mode functions (IMFs). An IMF must meet the following conditions: the number of endpoints and the number of crossed zeros must be equal to or differ by at most one; and the average value of an envelope determined by the maximum location and an envelope determined by the minimum location is zero.

An MFI represents a single oscillatory mode similar to a component in a single harmonic function, thus making it possible to effectively decompose an input signal using the EMD method.

More specifically, in order to extract an MFI from an input signal s (t), an upper envelope may be produced by connecting all local endpoints terminated by the maximum location of the input signal s (t) using a method. mathematically defined cubic curve interpolation control nodes, and a lower envelope can be produced by connecting all local extreme points determined by the minimum location of the input signal s (t) using the mathematically cubic curve interpolation method. defined by control nodes. All values that the input sign s (t) can have can be between the upper envelope and the lower envelope.

Subsequently, the mean value m (t) of the upper envelope and the lower envelope can be calculated. Subsequently, a first component Ii1 (t) can be calculated by subtracting the mean value m (t) from the input signal s (t), as indicated by Equation (10):

Mathematical value 10

s (t) -m1 (t) = h1 (t)

If the first component hi (t) does not satisfy the above conditions of the MFI, it can be determined that the first component h] _ (t) is equal to the input signal s (t), and the above operation can be performed again until a first Ci (t) from the MFI meeting the above mentioned MFI conditions is obtained.

Since the first MFI C1It) is obtained, a residue ri (t) is obtained by subtracting the first MFI C1 (t), as indicated by Equation (11):

Mathematical value 11

s (t) -c1 (t) = r1 (t)

Subsequently, the aforementioned MFI extraction operation can be performed again using residue! (T) as the new input signal, thereby obtaining a second MFI C2 (t) and a residue r2 (t).

If a residue rn (t) obtained during the above mentioned IMF extraction operation has a constant value, either a monotonically increasing function or a simple period function with only one extreme point or no extreme point, the above-mentioned extraction operation IMF can be terminated.

As a result of the aforementioned MFI extraction operation, the input signal s (t) can be represented by the sum of a plurality of MFI Co (t) to Cm (t) and a final residue rm (t) as indicated. by Equation (12):

Mathematical value 12

<formula> formula see original document page 30 </formula>

where M indicates the total number of extracted MFIs.

The final residue rm (t) may reflect the general characteristics of the input signal s (t).

Figure 10 illustrates eleven MFIs and a final residue obtained by decomposing an original input signal using the EMD method. Referring to Figure 10, the frequency of an MFI obtained from the original input signal at an early stage of MFI extraction is higher than the frequency of an MFI obtained from the original input signal at a later stage of MFI extraction.

Extraction of the MFI can be simplified by using a SD standard deviation between a previous residue hl (k-D and a current residue hlk, as indicated by Equation (13):

Mathematical value 13

<formula> formula see original document page 30 </formula> If the SD standard deviation is less than a reference value, for example 0.3, the current residue hlk can be considered as an MFI.

In the meantime, a signal x (t) can be transformed into an analytical signal by the Hilbert transform, as indicated by Equation (14):

Mathematical value 14

<formula> formula see original document page 31 </formula>

where (t) indicates an instantaneous amplitude, (t) indicates an instantaneous phase, and H {} indicates theHilbert transform.

As a result of the Hilbert transform, an input signal can be converted into an analytical signal that consists of a real component and an imaginary component.

By applying the Hilbert transform on a signal with a mean of 0, frequency components can be obtained that can provide high resolution for both the time domain and the frequency domain.

In the following, it will be described in detail how the determining unit 112 illustrated in Fig. 4 determines which of a plurality of coding units should be used to encode each of a plurality of split signals obtained by decomposing an input signal.

Determination unit 122 can determine which speech encoder and audio encoder can most efficiently decode each of the split signals. In other words, determination unit 122 can decode to encode split signals that can be efficiently encoded by a speech encoder using any of the first to mth coding units 210 and 220 in a speech coder and deciding to encode split signals which can be efficiently encoded by an audio coder using any of the first to mrth coding units 210 and 220 in an audio encoder.

In the following, it will be described in detail how determining unit 122 determines which of a deflector encoder and an audio encoder can most efficiently encode a divided signal.

The determining unit 122 can measure the change in a split signal and can determine that the split signal can be encoded more efficiently by a speech coder than by an audio coder if the result of the measurement. is greater than a predefined reference value.

Alternatively, the determining unit 122 may measure a tonal component included in a certain part of a split signal and may determine that the split signal may be encoded more efficiently by an audio encoder than by a speech encoder if the measurement result is greater than one. preset reference value.

Figure 11 is a block diagram of one embodiment of the determination unit 122 illustrated in Figure 5. Referring to Figure 11, a determination unit includes a speech coding / decoding unit 500, a first filter bank 510, a second filter bank520, a determination unit 530 and a psychoacoustic modeling unit 540.

The unit of determination illustrated in Fig. 11 can determine which of a speech encoder and an audio encoder can most efficiently encode each divided signal.

Referring to Fig. 11, an input signal is coded by the speech coding / decoding unit 500, and the coded signal is decoded by the speech coding / decoding unit 500, thereby restoring the speech signal. original entry. The speech encoding / decoding unit 500 may include a multi-rate broadband speech encoder / decoder (AMR-WB), and the AMR-WB speech encoder / decoder may have a code-stimulated linear prediction (CELP) structure.

The input signal may be sampled downstream before being inserted into the speech coding / decoding unit 500. A signal transmitted by the speech coding / decoding unit 500 may be ascended, thereby restoring the input signal. input signal.

The input signal may be subjected to frequency conversion by the first filter bank 510.

The signal transmitted by speech coding / decoding unit 500 is converted to a frequency domain signal by the second filter bank 520. The first filter bank 510 or the second filter bank 520 may perform cosine transform, for example. for example, modified discrete transform (MDCT) into a signal inserted therein.

Both a frequency component of the original input signal transmitted by the first filter bank 510 and a frequency component of the restored input signal transmitted by the second filter bank 520 are inserted into the determining unit 530. Determination 530 can determine which of a speech encoder and an audio encoder can encode the input signal most efficiently based on the frequency components inserted therein.

More specifically, determining unit 530 can determine which of a speech encoder and an audio encoder can most efficiently encode the input signals based on the frequency components nelain by calculating the PE ± perceptual entropy of each of the frequency components, using the Equation (15):

Mathematical value 15

<formula> formula see original document page 34 </formula>

on what

<formula> formula see original document page 34 </formula> where x (j) indicates a coefficient of a frequency component, j indicates an index of the frequency component, indicates the size of the quantization step, nint () is a function that returns the nearest integer of its argument, and jlow <i> and jhigh (i) are an initial frequency index and an end frequency index, respectively, of a scale factor band.

The determining unit 530 can calculate the perceptual entropy of the frequency component of the original input signal and the perceptual entropy of the frequency component of the restored input signal using Equation (15), and can determine which of an audio encoder and deum Speech encoder is most efficient for use in coding the input signal based on the calculation results.

For example, if the perceptual entropy of the frequency component of the original input signal is lower than the perceptual entropy of the frequency component of the restored input signal, the determining unit 530 may determine that the input signal may be more efficiently encoded by an audio encoder than by a speech encoder. On the other hand, if the perceptual entropy of the restored input signal frequency component is less than the perceptual entropy of the original input signal frequency component, the determining unit 530 may determine that the input signal frequency The input can be encoded more efficiently by a speech encoder than by an audio encoder. Figure 12 is a block diagram of a modality from one of the first to the seventh coding units 210 and 220 illustrated in Figure 1. The encoding code illustrated in FIG. 12 may be a speech encoder.

In general, speech coders can perform LPC on an input signal in frame units and extract a coefficient from LPC, for example, an order 16PC LPC coefficient, from each input signal frame using the Levinson-Durbin algorithm. A stimulus signal may be quantized by means of an adaptive codebook search or a fixed codebook search. The stimulus signal can be quantized using an algebraic code-stimulated linear prognostic method. Vector quantization can be performed on stimulus signal gain using a quantization table with a conjugate structure.

The speech coder illustrated in FIG. 12 includes a linear prediction analysis unit 600, a spacing estimation unit 610, a codebook debug unit 620, a spectrolinear pair (LSP) unit 630 and a quantization 640.

Linear prognostic analysis unit 600 performs linear prognostic analysis on an input signal that uses an autocorrelation coefficient that is obtained using an asymmetric window. If an anticipation period, that is, the asymmetric window, has a length of 30 ms, then the linear prediction analysis unit 600 can perform the linear prediction analysis using a 5 ms anticipation period. The autocorrelation coefficient is converted to a linear prognostic coefficient using a Levin-son-Durbin algorithm. For quantization and linear interpolation, the LSP 630 unit converts the linear prognostic coefficient into an LSP. Quantization unit 64 quantizes the LSP.

Offset estimation unit 610 is open loop offset estimation in order to reduce the complexity of an adaptive codebook search. More specifically, the offset estimation unit 610 estimates an open loop offset period using a weighted desinal speech domain of each frame. Subsequently, a harmonic noise modeling filter is configured using the estimated open loop offset. Subsequently, an impulse response is calculated using the harmonic noise modeling filter, a li-near prognostic synthesis filter, and a perceptual acoustic frequency peak weighting filter. The impulse response can be used to generate a target signal for the quantization of a stimulus signal.

The codebook search unit 620 performs an adaptive codebook search and a fixed code book search. The adaptive codebook search can be performed in subframe units by calculating an adaptive codebook vector by means of a closed loop spacing search and by interpolating the anterior stimulus signals. The stimulus signal can be generated by a linear prognostic synthesis filter to simplify a closed loop search. A fixed codebook structure is established based on the interlaced single pulse permutation (ISSP) design. A codebook vector comprising 64 positions, where 64 pulses are respectively located, is divided into four tracks, each track comprising 16 positions. A predetermined number of pulses can be located on each of the four tracks according to the transmission rate. Since a codebook index indicates the location of the track and the signal of a pulse, there is no need to store a codebook, and a stimulus signal can be generated simply by using the book index. code.

The speech coder illustrated in FIG. 12 may perform the above coding processes in a temporal domain. Also, if an input signal is encoded using a linear prognostic coding method by the classification module 100 illustrated in Figure 1, the linear prognostic analysis unit 600 may be optional.

The present invention is not restricted to the speech encoder illustrated in FIG. 12. In other words, various speech encoders other than the i-polished speech encoder in FIG. 12, which can encode speech signals efficiently, may be used within the scope of the present invention. .

Fig. 13 is a block diagram of another mode from one of the first to mrth coding units 210 and 220 shown in Fig. 1. The coding unit shown in Fig. 13 may be an audio coder.

Referring to Figure 13, the audio encoder includes a first filter bank 700, a psychoacoustic modeling unit 710, and a quantization unit 720.

Filter bank 700 converts an input signal into a frequency domain signal. Filter bank 700 may perform cosine transform, for example modified discrete transform (MDCT), on the input signal.

The psychoacoustic modeling unit 710 calculates an input signal SMR input signal masking threshold. The quantization unit 720 quantifies the co-efficient MDCT transmitted by the filter bank 700 using the masking limit calculated by the psychoacoustic modeling unit 710. Alternatively, to minimize the audible distortion in a given range. bit rate, the quantization unit 720 can use the input signal SMR.

The audio encoder illustrated in FIG. 13 may perform the above coding processes on a frequency domain.

The present invention is not restricted to the audio encoder illustrated in Figure 13. In other words, various audio encoders (e.g., advanced audio encoders) other than the audio encoder illustrated in Figure 13, which can efficiently encode audio signals, can be used in the scope of the present invention.

Advanced audio encoders perform temporal noise (TNS), intensity / coupling, prognosis and mid / side stereo coding modeling. TNS is an operation of properly distributing the time domain quantization noise in a filter bank window so that the quantization noise can be inaudible. Intensity / coupling is an operation that can reduce the amount of spatial information to be transmitted by encoding an audio signal and that can transmit audio signal energy only based on the fact that the perception of the direction of the Sound in a high band depends mainly on the temporary scale of the energy.

Prognosis is an operation of removing redundancy from a signal whose statistical characteristics do not vary by using the correlation between the frame spectral components. Stereo M / S encoding is an operation of transmitting the normalized sum (ie, middle) and the difference (ie, side) of a stereo signal instead of left and right channel signals.

A signal passing through TNS, intensity / coupling, prognosis, and M / S stereo coding is quantized by a quantizer performing Synthesis Analysis (AbS) using an SMR obtained from a psychoacoustic model.

As explained, since an audio encoder encodes an input signal using a modeling method, such as a linear prognostic coding method, the determination unit 122 illustrated in FIG. 5 can determine whether the audio signal is coded. Input can be easily modeled according to a predetermined set of rules. Subsequently, if it is determined that the input signal can be easily modeled, the determining unit 122 may decide to encode the input signal using a deflector encoder. On the other hand, if it is determined that the input signal cannot be easily modeled, the determining unit 122 may decide to encode the input signal using an audio encoder.

Figure 14 is a block diagram of a coding apparatus according to another embodiment of the present invention. In Figures 1 through 14, reference numerals are equal elements and thus their detailed descriptions will be omitted.

Referring to Figure 14, a rating module 100 splits an input signal in a plurality of the first to nth divided signals and determines which of a plurality of coding units 230, 240, 250, 260 and 270 should be used to encode each of the first to nth split signals.

Referring to Figure 14, coding units 230, 240, 250, 260 and 270 can sequentially encode the first to nth divided signals, respectively. Also, if the input signal is divided into a plurality of frequency band signals, the frequency band signals can be encoded in the order of a lower frequency band signal to a higher frequency band signal.

In a case where split signals are sequentially encoded, a coding error of a previous signal may be used to encode a current signal. As a result of this, it is possible to encode the divided signals using different encoding methods and thus prevent signal distortion and provide bandwidth scaling.

Referring to Figure 14, coding unit 230 encodes the first split signal, decodes the first coded split signal, and transmits an error between the decoded signal and the first divided signal to coding unit 240. Coding unit 240 it encodes the second split signal using the error transmitted by coding unit 230. In this way, the second to split mths are coded in consideration of the coding errors of their respective previous split signals. Therefore, you can perform encoding without error and improve sound quality.

The coding apparatus illustrated in FIG. 14 may restore a signal from a continuous stream of input bits by inversely performing the operations performed by the coding apparatus illustrated in FIGS. 1 through 14.

Figure 15 is a block diagram of a decoding apparatus according to one embodiment of the present invention. Referring to Figure 15, the decoding apparatus includes a bit unwrapping module 800, a decoder determination module 810, a decoding module 820 and a synthesizer module 830.

Bit unpacking module 800 extracts from a continuous stream of input bits one or more encoded signals and additional information that is required to decode the encoded signals.

The decoding module 820 includes a plurality of the first to mth decoding units 821 and 822 which perform different decoding methods.

The decoder determination module 810 terminates which of the first to mth decoding units 821 and 822 can most effectively decode each of the encoded signals. The decoder determination module 810 may use a method similar to that of the decoder module 100 illustrated in Figure 1 to determine which of the first to mth decoding units 821 and 822 can most efficiently decode each of the coded signals. In other words, the decoder determination module 810 can determine which of the first to most decoding units 821 and 822 can most effectively decode each of the encoded signals based on characteristics of each of the encoded signals. Preferably, the decoder determination module 810 can determine which of the first to mth decoding units 821 and 822 can most efficiently decode one of the encoded signals based on the additional information extracted from the input bit stream. .

Additional information may include information that identifies that a class in which a coded signal is classified belongs to a coding apparatus, coding unit information identifying a coding unit used to produce the coded signal, and information of the decoding unit that identifies a decoding unit to be used to decode the encoded signal.

For example, the decoder determination module 810 may determine which class an encoded signal belongs to based on the additional information and choose, for the encoded signals, any one of the first to the nth decoding units 821 and 822 corresponding to coded dosinal class. In this case, the chosen decoding unit may have a structure such that it can more efficiently decode signals belonging to the same encoded signal class.

Alternatively, the decoder determination module 810 may identify a coding unit used to produce a coded signal based on the additional information and to select for the coded signal any of the first to the most decoding units. 821 and 822 corresponding to the identified coding unit. For example, if the encoded signal was produced by a speech encoder, the decoder determination module 810 may choose, for the encoded signal, any of the first to mth decoding units 821 and 822 in a decoder Speech

Alternatively, the decoder determination module 810 may identify a decoding unit that can decode a coded signal based on further information and may choose, for the coded signal, any of the first to mth units. 821 and 822 corresponding to the i-dentified decoding unit.

Alternatively, the decoder determination module 810 may obtain the characteristics of a coded signal from the additional information and may choose any one of the first to mighty decoding units 821 and 822 that can most efficiently decode signals with the same characteristics as the coded signal.

In this manner, it is determined that each of the encoded signals extracted from the input bit stream that is encoded by any one of the first up to m-decoding units 821 and 822 may more efficiently decode a corresponding encoded signal. The decoded signals are synthesized by the synthesizer module 830, thereby restoring an original signal.

Bit unpacking module 800 extracts division information by considering the encoded signals, for example, the number of encoded signals and information from each of the encoded signals, and the synthesizing module 830 can synthesize the decoded signals provided by the decoding module 820 in relation to the division information.

The synthesizer module 830 may include a plurality of the first to nth synthesizer units 831 and 832. Each of the first to nth synthesizer units 831 and 832 may synthesize the decoded signals provided by the decode module 820 or You can perform additional domain conversion or decoding on the same or all of the encoded signals.

One of the first through nth synthesizing units 831 and 832 may perform a postprocessing operation, which is the inverse of a preprocessing operation performed by a coding apparatus, on a synthesized signal. Information indicating whether a postprocessing operation should be performed and decoding information used to perform the postprocessing operation can be extracted from the continuous stream of input bits.

Referring to Figure 16, one of the first to nth synthesizer units 831 and 832, particularly a second synthesizer unit 833, may include a plurality of the first to nth postprocessors 834 and 835. The first synthesizer unit 831 synthesizes a plurality of decoded signals into a single signal, and one of the first through nth postprocessors 834 and 835 performs a postprocessing operation on the single signal obtained by the synthesizers.

Information that indicates which of the first through nth postprocessors 834 and 835 should perform a postprocessing operation on the single signal obtained by synthesization can be included in the continuous stream of input bits.

One of the first to nth synthesizers 831 and 832 can perform linear prognostic decoding on the single signal obtained by the synthesizer unit using a linear prognostic co-efficient extracted from the input bit stream, thereby restoring an original signal. The present invention may be embodied as computer code readable written on a computer readable recording medium. Computer readable recording media can be any type of recording device in which data is stored in a computer readable manner. Examples of computer readable recording media include a ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage, and carrier (for example, data transmission over the Internet). Computer readable recording media may be distributed on a plurality of computer systems connected to a network so that computer readable code is written to it and executed from it in a decentralized manner. Functional programs, code and code segments necessary to carry out the present invention can be readily understood by those skilled in the art.

While the present invention has been particularly shown and described with respect to its exemplary embodiments, those skilled in the art understand that various changes in shape and detail can be made without departing from the spirit and scope of the present invention defined by the following. -few claims.

INDUSTRIAL APPLICABILITY

As stated, according to the present invention, it is possible to encode signals with different characteristics at an optimal bit rate by classifying the signals into one or more classes according to the characteristics of the signals and it is possible to encode each signal using a coding unit that can well serve the class to which a corresponding sign belongs. Therefore, it is possible to efficiently encode various signals, including audio and speech signals.

Claims (20)

1. Decoding method, characterized in that it comprises: extracting a plurality of encoded signals from a continuous stream of input bits, determining which of a plurality of decoding methods should be used to decode each of the encoded signals; signals encoded using the determined decoding methods; synthesize the decoded signals. The decoding method according to claim 1, characterized in that it further comprises extracting the information from the decoding method which considers how to decode each of the encoded signals, wherein the determination comprises determining why of the plurality of the methods. The decoded signals must be decoded using the information of the decoding method. 3. The decoding method according to claim 1, characterized in that the information of the decoding method comprises at least one of the information of the encoding unit that identifies a coding unit that produced a coded signal, of which decoding unit information identifying a decoding unit which shall decode the encoded signal and the information indicating a characteristic of the encoded signal. 4. Decoding method according to claim 1, characterized by the fact that determination comprises choosing any of the decoding methods which can most effectively decode each of the encoded signals. 5. A decoding method according to claim 1, characterized in that it further comprises extracting splitting information from the encoded signals of the continuous stream of input bits, wherein the synthesizing comprises synthesizing the signals. -decoded into a single signal with respect to division information. 6. The method of decoding according to claim 5, characterized in that the split information comprises numerous coded signals or frequency band information of the coded signals. A decoding method according to claim 1, characterized in that it further comprises extracting bit-quantity information from encoded signals from the continuous stream of input bits, wherein decoding comprises decoding. signals encoded according to the bit quantity information. A method of decoding according to claim 1, characterized in that it further comprises extracting encoded order decoding information from the continuous stream of input bits, wherein decoding comprises decoding encoded signals from according to the order-information of the decoding. A decoder apparatus, characterized in that it comprises: a bit unwrapping module that extracts a plurality of encoded signals from a continuous stream of input bits, a decoder determining module which determines which of a plurality of input units. The decoding must be used to decode each of the encoded signals, a decoding module comprising the decoding unit and decoding each of the coded signals using the determined decoding units, and a synthesizing module which synthesizes the decoding units. decoded signals. Decoding apparatus according to claim 9, characterized in that the bit-decoding module extracts information from the decoding unit of each of the encoded signals from the continuous stream of input bits, wherein the module The decoding determination parameter determines by which of the plurality of decoding units the encoded signals are to be decoded using the decoding unit information. 11. Decoder according to claim 9, characterized in that the decoder determination module chooses any of the decoding units which can most effectively decode the encoded signals. 12. A decoding apparatus according to claim 9, characterized in that the bit-wrapping module extracts coded signal division information from the continuous stream of input bits, wherein the synthesizer module synthesizes the decoded signals into one. single signal in relation to the division information. A coding method, characterized in that it comprises: dividing an input signal into a plurality of divided signals, determining which of a plurality of decoding methods should be used to encode each of the divided signals based on the characteristics of each of the divided signals; encode the divided signals using decoding methods; and generate a continuous stream of bits using the coded split signals. A coding method according to claim 13, characterized in that the determination comprises choosing any of the coding methods which can most efficiently encode the divided signals. An encoding method according to claim 13, characterized in that it further comprises allocating a number of bits to encode each of the divided signals. A coding method according to claim 13, characterized in that it further comprises determining an order in which the divided signals are to be coded. A coding method according to claim 13, characterized in that it further comprises splitting the input signal into a plurality of split signals, determining again which coding method is to be used for coding. each of the split signals, again determine a number of bits to encode the split signals or an order in which the split signals must be encoded. A coding apparatus, characterized in that it comprises: a signal division module that divides an input signal into a plurality of divided signals, an encoder determining module which determines which of a plurality of coding units must be used to encode each of the divided signals: a coding module comprising coding units and coding the divided signals using the determined coding units; It is a bit wrapping module that generates a continuous stream of bits using the encoded split signals. Coding apparatus according to Claim 18, characterized in that the encoder determining module chooses any of the coding units that can most efficiently encode the split signals. Computer readable recording media, characterized by the fact that you have a program for performing the decoding method of any one of claims 1 to 8 or the encoding method of any of claims 13 to 17.